DCG-SQL: Enhancing In-Context Learning for Text-to-SQL with Deep Contextual Schema Link Graph

2025-05-26

AI Summary

Existing Text-to-SQL approaches over-rely on the intrinsic capabilities of large language models (LLMs), employing coarse-grained and semantically misaligned example retrieval, which leads to sharp performance degradation in smaller models. To address this, we propose a retrieval-generation framework grounded in a Deep Contextual Pattern Linking Graph (DCPLG). Our method introduces a fine-grained semantic relationship graph explicitly modeling associations between natural language questions and database schema elements, thereby decoupling schema understanding from SQL generation. Guided by this graph structure, we perform context-aware example retrieval and integrate schema-linking–enhanced in-context learning. On the Spider benchmark, our approach improves the execution accuracy of Llama 3.1-8B by 12.7%, substantially narrowing the performance gap between small and large models, while preserving state-of-the-art (SOTA) accuracy for large models and accelerating inference.

Technology Category

• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Text-to-SQL, which translates a natural language question into an SQL query, has advanced with in-context learning of Large Language Models (LLMs). However, existing methods show little improvement in performance compared to randomly chosen demonstrations, and significant performance drops when smaller LLMs (e.g., Llama 3.1-8B) are used. This indicates that these methods heavily rely on the intrinsic capabilities of hyper-scaled LLMs, rather than effectively retrieving useful demonstrations. In this paper, we propose a novel approach for effectively retrieving demonstrations and generating SQL queries. We construct a Deep Contextual Schema Link Graph, which contains key information and semantic relationship between a question and its database schema items. This graph-based structure enables effective representation of Text-to-SQL samples and retrieval of useful demonstrations for in-context learning. Experimental results on the Spider benchmark demonstrate the effectiveness of our approach, showing consistent improvements in SQL generation performance and efficiency across both hyper-scaled LLMs and small LLMs. Our code will be released.

Problem

• Improving Text-to-SQL performance with in-context learning
• Addressing performance drops in smaller LLMs for SQL generation
• Enhancing demonstration retrieval using schema link graphs

Innovation

• Constructs Deep Contextual Schema Link Graph
• Retrieves useful demonstrations effectively
• Improves SQL generation across LLMs